University of Rhode Island DigitalCommons@URI
Biological Sciences Faculty Publications Biological Sciences
2015 The ffecE t of Light Intensity on Prey Detection Behavior in Two Lake Malawi Cichlids, Aulonocara stuartgranti and Tramitichromis sp. Margot A.B. Schwalbe
Jacqueline F. Webb University of Rhode Island, [email protected]
Follow this and additional works at: https://digitalcommons.uri.edu/bio_facpubs
The University of Rhode Island Faculty have made this article openly available. Please let us know how Open Access to this research benefits oy u.
This is a pre-publication author manuscript of the final, published article.
Terms of Use This article is made available under the terms and conditions applicable towards Open Access Policy Articles, as set forth in our Terms of Use.
Citation/Publisher Attribution Schwalbe MA, Webb JF. (2015). "The Effect of Light Intensity on Prey Detection Behavior in Two Lake Malawi Cichlids, Aulonocara stuartgranti and Tramitichromis sp." Journal of Comparative Physiology A: Neuroethology, Sensory, Neural, and Behavioral Physiology. Available at: http://link.springer.com/article/10.1007/s00359-015-0982-y#
This Article is brought to you for free and open access by the Biological Sciences at DigitalCommons@URI. It has been accepted for inclusion in Biological Sciences Faculty Publications by an authorized administrator of DigitalCommons@URI. For more information, please contact [email protected].
1
2 The effect of light intensity on prey detection behavior in two Lake Malawi cichlids,
3 Aulonocara stuartgranti and Tramitichromis sp.
4
5
6
7
8 Margot A. B. Schwalbe1 and Jacqueline F. Webb1
9
10 1Department of Biological Sciences, University of Rhode Island, 120 Flagg Road, Kingston, RI
11 02881, USA
12
13
14 Author for Correspondence: Margot A. B. Schwalbe, Department of Biology, Tufts University,
15 200 Boston Ave, Ste. 4700, Medford, MA USA; [email protected],
16 [email protected], telephone: (617) 627-0558, fax: (617) 627-0309
17
18
19 Tables: 5
20 Figures: 7 (2 color)
21
22
23
1
24 Abstract
25
26 Two Lake Malawi sand-dwelling cichlids (Aulonocara stuartgranti, Tramitichromis sp.) that
27 have different lateral line phenotypes, but feed on benthic invertebrates, have been shown to use
28 lateral line and/or visual cues to detect prey under light versus dark conditions. The current study
29 examined how ecologically relevant variation in light intensity (0-800 lux) influences detection
30 of prey (mobile, immobile) in each species by analyzing six well-defined behavioral parameters.
31 Both species fed at light intensities ≥1 lux; prey type and/or time of day (but not light intensity)
32 predicted all four parameters analyzed with generalized linear mixed models in A. stuartgranti,
33 whereas the interaction of light intensity and time of day predicted three of these parameters in
34 Tramitichromis sp. Data for all six parameters suggest that the critical light intensity is 1-12 lux
35 for both species, the integration of visual and lateral line input explains differences in the
36 detection of mobile and immobile prey and the behavioral changes that occur at the transition
37 from 1 to 0 lux in A. stuartgranti, and that Tramitichromis sp. likely uses binocular vision to
38 locate prey. The sensory biology of species that exploit similar food resources will have
39 important implications for the trophic ecology of African cichlid fishes.
40
41 Keywords
42 Vision, lateral line, detection distance, prey detection, sensory ecology
43
44 Abbreviations
45 AICC Akaike information criterion
46 GLMM Generalized linear mixed model
2
47 Lx Lux
48 PAR Photosynthetically active radiation
49 SL Standard length
50 TL Total length
51
52 Introduction
53
54 Light in aquatic habitats varies in quality and quantity over time and space (Kirk 2011) and
55 influences the ability of visual predators to detect and capture mobile prey (Vinyard and O’Brien
56 1976; Confer et al. 1978; Lythgoe 1979; Ryer and Olla 1999; Vogel and Beauchamp 1999;
57 Rickel and Genin 2005). Fishes occupying similar habitats may demonstrate variation in
58 visually-mediated prey detection abilities, such as visual thresholds and absorption spectra of
59 visual pigments, which may provide a competitive advantage under particular light conditions
60 (Vogel and Beauchamp 1999; Hofmann et al. 2009). Many fishes are also able to detect prey at
61 low light intensities (e.g., dawn, dusk, at depth, or with increased turbidity), but with reduced
62 capabilities compared to that at higher light intensities. The distance at which free swimming
63 prey are detected dramatically decreases below a certain light intensity (“critical light intensity,”
64 Confer et al. 1978) in salmonids (Dunbrack and Dill 1984; Henderson and Northcote 1985) and
65 some freshwater percomorphs (bluegill, Lepomis macrochirus, Vinyard and O’Brien 1976;
66 largemouth bass, Micropterus salmoides, Howick and O’Brien 1983; yellow perch, Perca
67 flavescans, Richmond et al. 2004).
68 Given the importance of multimodal sensory integration in the formulation of behavior,
69 the contributions of the non-visual sensory systems to prey detection (e.g., mechanosensory
3
70 lateral line, auditory, olfactory, gustatory, somatosensory/tactile, and in some cases, the
71 electrosensory system; reviewed in Montgomery et al. 2014) must also be considered.
72 Morphological and/or physiological specializations of non-visual sensory systems, including the
73 olfactory system (Parzefall 1993; Montgomery et al. 1999), gustatory system (Atema 1971) and
74 the lateral line system (Janssen 1997; Schwalbe et al. 2012, reviewed in Webb 2014), have been
75 used to predict how these senses provide alternatives to vision for prey detection in light-limited
76 environments. Futhermore, the integration of different combinations of sensory inputs may
77 explain variation in behavior under different environmental conditions (Partridge and Pitcher
78 1980; Moller 2002; Montgomery et al. 2003; Gardiner and Motta 2012). Several species of fishes
79 have been shown to modulate feeding strategies using a combination of visual and non-visual
80 cues that allow them to feed under a range of light conditions, including darkness (Townsend and
81 Risebrow 1982; Batty et al. 1986; Diehl 1988; Schwalbe et al. 2012).
82 The mechanosensory lateral line system is known to play important roles in prey
83 detection, as well as in predator avoidance, communication, and navigation around obstacles
84 (Webb et al. 2008; Montgomery et al. 2014). The system demonstrates a great deal of variation,
85 which is defined by the morphology of the cranial and trunk lateral line canals and neuromast
86 receptor organs within them, and the distribution of superficial neuromasts on the skin of the
87 head, trunk and tail (reviewed in Webb 2014). Widened lateral line canals, one of five cranial
88 lateral line canal phenotypes found among bony fishes, has evolved convergently in ~12 teleost
89 families (including deep sea taxa) and appears to be an adaptation for enhanced sensitivity to
90 water flows and prey detection (Denton and Gray 1988, 1989; Montgomery and Coombs 1992;
91 discussed in Schwalbe et al. 2012; reviewed in Webb 2014).
4
92 The speciose cichlid fishes of the African Rift Lakes are typically described as visual
93 feeders (Fryer and Iles 1972) and most genera have narrow cranial lateral line canals, but all
94 members of a few genera (e.g., Alticorpus, Aulonocara, Aulonocranus, Trematocara,
95 Trematocranus, Konings 2007) have widened lateral line canals suggesting the capacity for
96 lateral line mediated prey detection (Konings 1990). Two genera of non-mbuna, haplochromine
97 cichlids in Lake Malawi, Aulonocara (widened canals) and Tramitichromis (narrow canals; Fig.
98 1), provide an interesting taxon pair for comparison of prey detection strategies since both feed
99 on benthic invertebrates in the sand, and thus appear to be ecologically similar. Schwalbe et al.
100 (2012) and Schwalbe and Webb (2014) analyzed the behavioral responses of Aulonocara
101 stuartgranti and Tramitichromis sp. to tethered live and dead prey (=adult brine shrimp, Artemia
102 sp.), in experiments carried out under light and dark conditions in which the lateral line system
103 was experimentally inactivated. These studies demonstrated that A. stuartgranti uses a
104 combination of inputs to its visual and lateral line systems to detect prey in the light, but depends
105 on its lateral line system to detect prey in the dark. Furthermore, these studies showed that
106 deactivation of the lateral line system of A. stuartgranti significantly affected prey detection
107 behavior and revealed that other senses (olfaction, gustation, and somatosensory/tactile) were
108 insufficient to initiate prey detection behavior in the dark. In contrast, Tramitichromis sp. did not
109 feed in the dark, and the inactivation of the lateral line system had little effect on prey detection
110 behavior in the presence of light, demonstrating that it is a visual predator.
111 Aulonocara and Tramitichromis species appear to share a food resource (benthic
112 invertebrates in sandy substrates), but occupy different depth ranges (Aulonocara species at
113 depths of 5-120 m and Tramitichromis species at depths of < 15 m; Fryer and Iles 1972; Konings
114 1990, 2007) and use different strategies to detect and capture benthic invertebrate prey in the
5
115 field. Species of Aulonocara swim just above the substrate to sense hydrodynamic flows
116 generated by benthic invertebrates in or on the substrate in the field (Konings 2007). In contrast,
117 species of Tramitichromis typically capture invertebrate prey by plunging into the substrate,
118 filling their mouth with sand, and sifting out prey with their gill rakers in the field (= “sand
119 sifting,” Fryer, 1959). This sand sifting behavior appears to be synonymous with the
120 “winnowing” behaviors observed in some surfperches (Laur and Ebeling 1983) and vision likely
121 contributes to the ability to locate patches of high quality food resources where “winnowing”
122 takes place (Holbrook and Schmitt 1984; Schmitt and Holbrook 1984).
123 This study uses the same methods used in prior studies (Schwalbe et al. 2012; Schwalbe
124 and Webb 2014) to test the hypothesis that variation in light intensity (0-800 lx) will have
125 different effects on the detection of live (mobile) and dead (immobile) benthic invertebrate prey
126 in A. stuartgranti and Tramitichromis sp.
127
128 Materials and methods
129
130 Study species
131
132 Adult Aulonocara stuartgranti and Tramitichromis sp. (unidentifiable to species level, J.
133 Stauffer, pers. commun.), is referred to as Tramitichromis throughout. These fish were reared in
134 the laboratory from breeding stock originally acquired from commercial suppliers (A.
135 stuartgranti: Bluegrass Aquatics, Louisville, KY, USA; Tramitichromis: Old World Exotic Fish,
136 Inc., Homestead, FL, USA and Life Fish Direct, Draper, UT, USA). They were housed in small
137 groups by species in 190 L aquaria at 26±1°C and 1.0±0.2 p.p.t. salinity (using Cichlid Lake Salt,
6
138 Seachem Laboratories, Inc., Madison, GA, USA) with standard white fluorescent light on a
139 12h:12h diurnal cycle and, equipped with appropriate mechanical and biological filtration. Fish
140 were fed daily with cichlid pellets (New Life Spectrum Cichlid Formula; New Life International,
141 Inc., Homestead, FL) and supplemented with live adult brine shrimp. Individual fish were not
142 used in feeding experiments if breeding behavior was observed. Animal care and all
143 experimental procedures followed an approved IACUC protocol.
144
145 Light environment in the experimental tank
146
147 Light in the experimental tank was provided by two fluorescent light fixtures (Lithonia Lighting,
148 Model GRW 2 14 CSW CO M4, Conyers, GA, USA) fitted with full spectrum bulbs (BlueMax
149 lamps, Full Spectrum Solutions, Jackson, MI, USA) positioned above the tank and within an
150 opaque curtain enclosure. The curtain (black canvas) was suspended from a rectangular plywood
151 frame placed 2 m above the top of the tank in order to exclude ambient light from entering the
152 set-up during all behavioral trials (Fig. 2a). Light intensity was varied by changing the height of
153 the lights above the water surface and using combinations of different neutral density filters
154 covering the lights (Lee Filters, Burbank, CA, USA). Light intensity (in lux [lx], lumen/m2, and
155 photosynthetically active radiation [PAR], µmol photons/m2/s) and color spectrum were
156 measured using a spectrometer (range: 340-1028 nm, Jaz spectrometer, Ocean Optics, Dunedin,
157 FL, USA) connected to a 2 m optical fiber (QP400-2-UV/VIS, Ocean Optics) fitted with a cosine
158 corrector (CC-3, Ocean Optics). Water temperature was monitored during experiments and the
159 fluorescent bulbs did not raise the temperature of the experimental tank.
7
160 Light intensities used in this study were based on the following data and calculations.
161 First, light levels present during sunrise/sunset to darkness are known for other freshwater
162 habitats (Harden Jones 1956; Ali 1959) and can range from 1000 lx (early twilight) to 0 lx (new
163 moon, Table 1). Second, few direct measurements of light intensities at different depths in Lake
164 Malawi are available, so the light intensity at specific depths were estimated with the following
165 equation:
!! ×! �! = �!× �
166 where Is and It are the light intensities at the surface (S) and at depth (T); and ε is the light
167 extinction coefficient. The average light intensity at the surface of Lake Malawi at midday on a
168 clear sunny day is approximately 2000 µmol photons/m2/s (~108,000 lx). This photon flux was
169 derived from cloudless surface irradiance for Lake Malawi (Guildford et al. 2000). Using light
170 extinction coefficient of either 0.10 m-1 (Patterson et al. 2000), 0.13 m-1 (Guildford et al. 2007),
171 or 0.43 m-1 (Guildford et al. 2007) depending on location and season, the light intensity at many
172 depths can be estimated under these conditions (Table 1).
173 Full spectrum bulbs were used because they provide the range of wavelengths that
174 correspond to the range of known absorption peaks of retinal photopigments in species of
175 Aulonocara and Tramitichromis. For instance, absorption peaks for A. hueseri are at 415 nm
176 (violet), 484 nm (blue-green) and 526 nm (green; Jordan et al. 2006) and absorption peaks for T.
177 intermedius are at 455 nm (blue), 532 nm (green) and 569 nm (red; Parry et al. 2005). In the
178 experimental tank, full spectrum bulbs generated major and minor light peaks at 404, 435, 487,
179 545, 587, and 611 nm, and neutral density filters were used to change light intensity did not
180 appreciably change the light spectrum in the experimental tank (Fig. 2b)
181
8
182 Experiments
183
184 Behavioral trials and video analysis of six well-defined behavioral parameters (number of prey
185 strikes, detection distance, detection angle, detection-to-strike velocity, swimming phase [glide,
186 pause] at detection and, prey type preference [order of prey strikes]) were carried out as in
187 Schwalbe et al. (2012) and Schwalbe and Webb (2014) with slight modifications. A total of sixty
188 trials were conducted using A. stuartgranti (30 trials, n = 6 fish, 75-85 mm total length [TL], 4
189 females, 2 male) and Tramitichromis (30 trials, n = 6 fish, 75-98 mm TL, 1 female, 5 males) in
190 order to quantify variation in behavioral responses to live (mobile) and dead (immobile) prey (=
191 tethered adult brine shrimp) at five light intensities between 0 and 800 lx.
192 Trials were conducted in an experimental tank (120 x 75 x 60 cm; 560 L) with 5 cm of
193 sand covering the bottom of the tank. Light intensity and spectral measurements (with ±0.01
194 accuracy, measured in lx and PAR) were taken directly above the center of each mesh platform
195 (to which live and dead prey were tethered, see below) before and after each trial, and light
196 intensity and spectrum were found to be consistent at all six platforms and trials (Figs. 2b, c).
197 Each fish was acclimated to a particular light intensity in the experimental tank for at least 30
198 minutes prior to a trial. The transition between photopic (cone-mediated) and scotopic (rod-
199 mediated) vision occurs at approximately 1 lx, and light-adapted fish may take 30 minutes (and
200 up to 3 hours) to become dark-adapted (Ali 1959). Thus, the 30+ minute light adaptation period
201 was judged to be sufficient to allow the fish’s visual system to adjust to the light level for a given
202 trial.
203 Before each trial, 12 adult brine shrimp (Artemia sp.) were tethered in pairs (1 live and 1
204 dead, freshly frozen) on each of six mesh platforms (10 x 10 cm), which were positioned in a 2 x
9
205 3 matrix so that the top of each platform flush with the sand surface. The water filtration system
206 for the experimental tank was then turned off to eliminate acoustic and hydrodynamic noise. At
207 the start of a trial, a fish was released into the experimental arena from behind an opaque divider
208 and feeding behavior was recorded for 30 min using an HD digital video camera (Sony © HDR-
209 CX550V, 30 frames per second) mounted directly above the tank, which provided a dorsal view
210 of the experimental arena. Trials at 1-800 lx were carried out with standard fluorescent room
211 lights on for all but the lowest light levels (1-12 lx). Dark trials (0 lx) were conducted with room
212 lights off, but with infrared illumination (peak = 840 nm, range 800-880 nm; Speco Provideo,
213 IR-200/24, Amityville, NY) to allow video recording of behavior.
214 Each fish was run through five trials, one trial per day each at a single light intensity,
215 progressing from highest to lowest intensity on subsequent days (e.g., 800, 112, 12, 1, and then 0
216 lx). Trials were carried out in this order to increase the likelihood that a fish would respond to
217 prey at lower light intensities (especially in the dark, 0 lx), as was suggested by preliminary
218 results. Trials were conducted over four months and the mean time between the first (800 lx) and
219 last (0 lx) trial for a given fish was 11 days (range = 6-19 days).
220 “Light” trials (1-800 lx) started midday to late afternoon (11:00-17:00) and “dark” trials
221 (0 lx) took place shortly after sunset (19:00-21:00; soon after room lights had automatically shut
222 off; as in Schwalbe et al. 2012; Schwalbe and Webb 2014). Dark trials (0 lx) were not carried out
223 during the day (during the light phase of the lab’s light:dark cycle) in order to avoid the
224 introduction of extraneous light. In addition, it was known that placing fish in low light or
225 darkness during normal daylight hours would disrupt feeding behavior (M.A.B. Schwalbe and A.
226 Mensinger, pers. obs.), and that species that normally feed both in full light during the day and at
10
227 night (e.g., during the dark phase of a lab's light:dark cycle) were unresponsive in dark (0 lx)
228 trials carried out during the day.
229 To assess the number of prey detections that lead to prey strikes, unconsumed prey were
230 counted at the end of each 30-minute trial and strike success was also confirmed in video
231 recordings. Video sequences leading to each prey strike were exported to Premier Pro (Adobe,
232 CS5) for further analysis. Analysis of sequential video frames was used to identify the phase of
233 swimming behavior (thrust, glide, or pause) during which prey detections occurred. Detection
234 distance and detection angle were measured in these images using ImageJ (NIH, v. 1.41o).
235 Detection distance was defined as the distance from the tip of a fish’s mouth to the prey,
236 measured in the frame immediately before the fish oriented towards it (e.g. before a turn defining
237 detection). For each prey strike, detection-to-strike velocity was calculated by dividing detection
238 distance by the time interval between detection and initiation of a strike. Detection angle was
239 measured in the same video frame in which detection distance was measured, and was defined as
240 the angle between a line extending anteriorly along midline of the fish (body axis) and a line
241 drawn from the prey to the tip of the fish's mouth.
242
243 Statistical analysis
244
245 Four of the six behavioral parameters were analyzed using generalized linear mixed models
246 (GLMMs; number of prey strikes, detection distance, detection-to-strike velocity, phase of
247 search behavior during which detections occurred). In addition, a ranking method (Taplin 2007)
248 was used to analyze prey preferences (live versus dead prey) and circular statistics were used to
249 analyze detection angles. All continuous data (e.g. detection distance and detection-to-strike
11
250 velocity) were tested for normality (Kolmogorov-Smirnov test) and were log10 transformed to
251 achieve normality (detection distance and detection-to-strike velocity). All statistical tests were
252 considered significant at P < 0.05 and values are given as means ± s.e.m.
253 Start time (= time of day, 0-24 hr) for trials conducted at the five different light intensities
254 was analyzed with nonparametric tests (e.g. Kruskal-Wallis test and Mann-Whitney U test) to
255 determine whether time of day affected feeding behavior. This analysis showed that the times at
256 which light trials (1-800 lx) and dark trials (0 lx) started did not differ between species (Mann-
257 Whitney U test, P > 0.05), but trial start time varied among light intensities in each species
258 (Kruskal-Wallis test, A. stuartgranti: K = 22.804, P < 0.001; Tramitichromis: K = 20.141, P <
259 0.001). Thus, time of day (=trial start time) was included in all GLMM analyses.
260 Four-way GLMM analyses (SPSS, IBM, v. 22) were used to test whether species (A.
261 stuartgranti, Tramitichromis), light intensity (0-800 lx), prey type (live, dead), and/or trial start
262 time (0-24 hr) predict differences in each of four behavioral parameters (number of prey strikes,
263 detection distance, detection-to-strike velocity, and phase of search behavior during which
264 detections occurred). Three-way GLMM analyses were used to further examine whether light
265 intensity, prey type, and/or trial start time predict differences in the four behavioral parameters in
266 each species separately. The selection of random (individual) and fixed effects (species, light
267 intensity, prey type, and trial start time), including repeated measures for the same individual,
268 was addressed in all analyses. Different types of GLMMs were used to account for the different
269 types of data collected in this study (summarized in Table 2) and the most parsimonious model
270 was selected for each behavioral parameter based on the corrected Akaike information criterion
271 (AICC).
12
272 The order in which live (mobile) and dead (immobile) prey were struck was analyzed in
273 each species following Taplin (2007). This method assumes that when presented with equal
274 numbers of two or more types of prey, the order in which prey are consumed provides
275 information about prey preference – that prey consumed first are more highly preferred than prey
276 consumed second, third, etc. and the last prey consumed is the least preferred. While differences
277 in handling time, encounter rates, and relative mobility of prey can potentially complicate the
278 results of this sort of analysis (Durham et al. 2012, McWilliam et al. 2013), such variation was
279 minimized in the current study by offering equal numbers of live and dead brine shrimp tethered
280 in the same arrangement to platforms placed in a 2 x 3 matrix in all trials. The null hypothesis for
281 this analysis was that live and dead prey would be consumed randomly during a trial. Videos
282 were analyzed so that each prey consumed was assigned a rank number (first prey consumed=1,
283 second prey consumed=2, etc.), and any remaining prey were assigned an average of the
284 remaining preference scores, and considered “tied for last.” A pair of preference scores for live
285 and dead prey at each light intensity was calculated for each fish. The pairs of scores from all of
286 the fish were considered independent samples and thus grouped by light intensity and species for
287 analysis. A score of 6.5 (based on presentation of six live and six dead prey, 12 total prey in a
288 trial) indicated no preference, a score of <6.5 revealed a preference for that prey type, and a score
289 of >6.5 indicated no preference or that prey type was ignored or avoided. Scores for live and
290 dead prey at each light intensity and for each species, were compared separately using paired t-
291 tests (SPSS, IBM, v. 22).
292 Detection angles were analyzed with circular statistics (Oriana v. 3, Kovach Computing
293 Services, Anglesey, UK). Rayleigh tests were performed for each species to test whether
294 detections of live and dead prey at each of the five light intensities (0-800 lx) occurred at
13
295 uniformly or non-uniformly distributed positions around the fishes’ body relative to the body
296 axis (e.g. to define the receptive field). Watson’s U2 tests were used to determine if detection
297 angles differed with prey type and with light intensity within and between species.
298
299 Results
300
301 Aulonocara stuartgranti and Tramitichromis sp. actively swam around the experimental tank
302 during trials at all light intensities, including darkness (0 lx), and used a saltatory search strategy
303 (a cyclic sequence of a caudal fin thrust, glide, and pause) while exploring the tank. Of the 360
304 total prey presented to fish during all 60 trials, A. stuartgranti struck at 299 prey (=83%) and
305 Tramitichromis struck at 231 prey (=64%; see Figs. 3, 4). Prey were detected by both species
306 during a glide or a pause, but never during a thrust (see Fig. 5).
307 Four-way GLMM analyses (Table 3) indicated that species alone did not predict
308 differences in any of the four behavioral parameters (number of prey detections, detection
309 distance, detection-to-strike velocity, or swimming phase at prey detection), and that only the
310 interaction of species, light intensity, and prey type had a significant effect only on number of
311 prey detections. The interaction of light intensity and time of day predicted differences in all four
312 behaviors. Light intensity alone predicted differences in all four behaviors, and time of day
313 predicted differences in three behavioral parameters (number of prey detections, detection
314 distance, detection-to-strike velocity), but not in swimming phase at prey detection. Separate
315 three-way GLMM analyses for each species (Table 4, see below) revealed interesting trends that
316 are indicative of species differences in prey detection behavior. Analyses of prey preference (live
14
317 vs. dead prey) and prey detection angle, carried out using other statistical methods, also indicated
318 differences in behavior between species, but did not consider time of day.
319
320 Feeding behavior of Aulonocara stuartgranti
321
322 GLMM analyses (Table 4) showed that light intensity did not significantly predict any of the
323 four behavioral parameters in A. stuartgranti (number of prey detections, detection distance,
324 detection-to-strike velocity, or swimming phase at prey detection; GLMMs, P > 0.05; Table 4).
325 However, time of day predicted the number of prey detections, and the interaction of time of day
326 and prey type predicted both detection distance and detection-to-strike velocity (GLMMs, P <
327 0.05; Table 4). Neither light intensity, time of day, prey type, nor their interactions, predicted
328 swimming phase at prey detection.
329 An examination of data for each of the behavioral parameters revealed informative
330 trends. A. stuartgranti struck at high numbers of both live (mobile) and dead (immobile) prey at
331 light intensities of 1-800 lx (Figs. 3a, 4a). They detected prey during a pause about half of the
332 time, but detected 61% of prey in a pause at the highest light intensity (800 lx; Fig. 5a).
333 Detection distance and detection-to-strike velocity appeared to not vary among light intensities
334 of 12-800 lx, but fish tended to detect live prey from greater distances (mean = 8.7-9.6 cm) than
335 dead prey (6.0-6.9 cm) and to detect live prey at higher detection-to-strike velocities (9.7-10.4
336 cm/s) than dead prey (6.9-7.4 cm/s; Figs. 4c, e). At 1 lx, fish tended to detect live and dead prey
337 from similar distances (mean = 6.5 and 6.3 cm, respectively; Fig. 4c) and similar detection-to-
338 strike velocities (6.8 cm/s and 7.8 cm/s, respectively; Fig. 4e). Live and dead prey (combined)
339 were detected at non-uniformly distributed positions around the fishes’ bodies at light intensities
15
340 ≥1 lx (Rayleigh test, P < 0.001; ±90° from body axis) with no differences in the distribution of
341 angles among pairs of light intensities with the exception of the two highest light intensities (112
342 lx versus 800 lx; Watson’s U2 test, U2 = 0.19, P < 0.05; Fig. 6a). Finally, fish tended to prefer
343 live prey at all light intensities, but only demonstrated a statistically significant preference for
344 live prey at 112 lx, but not at 800 lx (Table 5; Fig. 7a), which is not easily explained.
345 In the dark (0 lx), prey detection behavior of A. stuartgranti was different than at light
346 intensities ≥ 1 lx. Fish struck at only 22 prey (=30.6% of total prey presented; Fig. 3a), and
347 tended to detect prey at even shorter distances (Fig. 4c) and at slower detection-to-strike
348 velocities (Fig. 4e) than when at least some light was present. Fish tended to detect more live
349 prey than dead prey (mean of 2.7 and 1.0, respectively; Figs. 3a, 4a), showed a statistically
350 significant preference for live prey (Table 5, Fig. 7a), and detected live prey from more than
351 twice the distance than dead prey (3.2 and 1.4 cm, respectively; Fig. 4c). In addition, detection-
352 to-strike velocity at 0 lx was about one half of that at higher light intensities (~3.5-5 cm/s at 0 lx
353 versus ~7-10 cm/s at ≥ 1 lx), but fish tended to detect live prey at somewhat higher detection-to-
354 strike velocities than dead prey (Fig. 4e). In the dark, 95% of prey were detected during a glide
355 and only a few prey (5%) were detected during a pause (Fig. 5a). Prey (live and dead combined)
356 were detected at positions uniformly distributed around fishes’ bodies (Rayleigh test, P > 0.05)
357 at a wide range of angles (± 180° from body axis, Fig. 6a), but the distribution of detection
358 angles did not differ for live versus dead prey (Watson’s U2 test, P > 0.05).
359
360 Feeding behavior of Tramitichromis
361
16
362 GLMM analyses (Table 4) showed that, in contrast to A. stuartgranti, the interaction of light
363 intensity and time of day predicted three of four behavioral parameters (number of prey
364 detections, detection distance, and detection-to-strike velocity). As in A. stuartgranti, neither
365 light intensity, time of day, nor prey type, or their interactions, predicted swimming phase at prey
366 detection. Prey type did not predict any of the four behavioral parameters in Tramitichromis, and
367 the interaction of light intensity and prey type predicted only detection distance (Table 4).
368 An examination of trends for each of the behavioral parameters revealed that
369 Tramitichromis tended to strike at high numbers of prey (Fig. 3b, 4b), and >60% of prey (live
370 and dead, combined) were detected during a pause at light intensities of 1-800 lx (Fig. 5b). At
371 light intensities of 12-800 lx, fish struck at live and dead prey from similar, long detection
372 distances (means = 9.8-10.1 and 8.5-10.0 cm, respectively) and at high detection-to-strike
373 velocities (9.6-10.5 and 8.7-9.3 cm/s, respectively). In contrast, at 1 lx, fish tended to strike at
374 both live and dead prey at similar, but shorter detection distances (6.9 and 6.3, respectively) and
375 lower detection-to-strike velocities (7.0 and 6.1, respectively; Figs. 4d, f) than at higher light
376 intensities. Both live and dead prey were detected at non-uniform positions around the body
377 (Rayleigh test, P < 0.001), which defined a very narrow range of detection angles from the body
378 axis (±40°); distributions were the same for live prey and dead prey at light intensities of 1-800
379 lx (Watson’s U2, P > 0.05; Fig. 6b). Fish tended to prefer live prey at different light intensities,
380 but only showed a statistically significant preference for live prey at the highest light intensity
381 (800 lx; Table 5; Fig. 7b).
382 Despite being active in the dark (0 lx), Tramitichromis only struck at only 3 prey (=4.2%
383 of the 72 prey presented). These strikes are likely to have been the result of random encounters
384 with prey as opposed to being the result of active search and directed strikes.
17
385
386 Discussion
387
388 The multiple statistical analyses presented here, and the detailed examination of trends in the
389 detection of live and dead prey at different light intensities in each species indicate that light
390 intensity affects prey detection behavior in different ways in Aulonocara stuartgranti and
391 Tramitichromis.
392
393 Feeding behavior of Aulonocara stuartgranti and Tramitichromis
394
395 Prey type and/or time of day, but not light intensity, were predictors of three of the four
396 behavioral parameters (number of prey detections, detection distance, and detection-to-strike
397 velocity) analyzed using GLMMs in A. stuartgranti. The lack of significance for light intensity is
398 consistent with the use of lateral line cues (see also Schwalbe et al. 2012), but also suggests that
399 A. stuartgranti may use a light-independent circadian rhythm to interpret time of day. This is
400 consistent with the occurrence of Aulonocara species at depths up to120 m in Lake Malawi
401 where light is limited or absent (Konings 1990, 2007) and in caves where spawning has been
402 reported (Grant et al. 1987), and thus where normal diurnal variation in light intensity may not be
403 a consistent or reliable cue for the regulation of behavior. The significance of prey type as a
404 predictor of detection distance and detection-to-strike velocity is illustrated by apparent
405 differences in numbers of live and dead prey detected at the same light intensities (Fig. 4; see
406 also Schwalbe et al. 2012; Schwalbe and Webb 2014), the tendency to prefer live prey at all light
407 intensities, and the statistically significant preference for live prey in the dark.
18
408 In contrast, in Tramitichromis, it is the interaction of light intensity and time of day that
409 predict these same three behavioral parameters. The importance of light intensity not surprising
410 because Tramitichromis uses visual, but not lateral line cues, for prey detection and does not feed
411 in the dark (Schwalbe and Webb 2014). Furthermore, these two factors are correlated both in the
412 lab where the fish were reared (on a 12:12 hr light/dark cycle) as well as in the relatively shallow
413 waters in their natural habitat in Lake Malawi, which is just 9-17° south of the equator where
414 fish experience 11-13 hours of daylight per day on an annual basis (http://astro.unl.edu). Thus,
415 these fish have evolved and are reared in environments where light intensity and time of day are
416 tightly correlated. The independent roles of these two factors in predicting behavior would need
417 to be addressed in additional experiments, which were out of the scope of this study.
418 Swimming phase (glide, pause) during which prey were detected was predicted neither
419 by light intensity nor by time of day in either species. The ability to detect prey during a glide or
420 pause will affect both the stabilization of the visual field (for vision-mediated detection) and/or
421 the magnitude of environmental and self-generated hydrodynamic noise (for lateral line-
422 mediated detection). A. stuartgranti and Tramitichromis both detected between 40% and 70% of
423 prey during a pause at light intensities of ≥1 lx, suggesting the importance of stabilizing the
424 visual field for prey detection at these light intensities. Prey type (which defines the presence or
425 absence of an additional visual motion stimulus) did not predict swimming phase at detection for
426 Tramitichromis (P < 0.053), but a larger sample size may have yielded a different statistical
427 outcome. Prey type also did not predict swimming phase at prey detection for A. stuartgranti, but
428 the shift to 95% of prey detections during a glide in the dark (where stabilization of visual field is
429 irrelevant), and their preference for live prey (that generate hydrodynamic flows detected in the
430 dark; Schwalbe and Webb 2014), are important indicators of the overall importance of prey type.
19
431 A. stuartgranti detected live prey at distances of less than half of a body length and at lower
432 detection-to-strike velocities at a low light intensity (1 lx) and in the dark (0 lx). In the presence
433 of at least some light, lower detection-to-strike velocities would also reduce self-generated
434 hydrodynamic noise (Montgomery et al. 2009), enhancing lateral line-mediated prey detection,
435 which would suggest that fish would tend to detect prey during a pause. However, the high
436 proportion of detections (95%) at relatively low detection-to-strike velocities, while not
437 eliminating self-generated noise, would bring a fish into the vicinity of potential prey that are
438 generating detectable hydrodynamic flows (Schwalbe et al. 2012).
439
440 Role of vision and critical light intensities
441
442 The importance of vision in A. stuartgranti and Tramitichromis is further supported by a
443 consideration of critical light intensities and the potential differences in the use of binocular
444 vision. Prey detection at relatively long distances is consistent with vision-mediated prey
445 detection in fishes (Vinyard and O’Brien 1976, Confer et al. 1978, Henderson and Northcote
446 1985, Mazur and Beauchamp 2003), and at higher light intensities detection of free swimming
447 prey generally occurs at longer distances (Vinyard and O’Brien 1976; Richmond et al. 2004;
448 Bergstrom and Mensinger 2009). In this study, both A. stuartgranti and Tramitichromis tended to
449 demonstrate the longest detection distances at the highest light intensities, which is thus
450 consistent with vision-mediated prey detection. Detection distances may not increase as light
451 intensity increases further in a given species (Schmidt and O’Brien 1982), but may decrease
452 sharply below a “critical light intensity” (Confer et al. 1978). Trends in behavioral parameters in
453 the current study reveals that the critical light intensity for fish feeding on tethered adult brine
20
454 shrimp is between 12 and 1 lx for both A. stuartgranti and Tramitichromis. This is comparable to
455 the critical light intensities for other freshwater teleosts in studies feeding on free-swimming
456 Daphnia (11-50 lx, in bluegill, Vinyard and O’Brien 1976; in lake trout, brook trout, and
457 bluegill, Confer et al. 1978), amphipods (5-25 lx, in round goby, logperch, slimy sculpin, and
458 spoonhead sculpin, Bergstrom and Mensinger 2009), or on small fish (~6-18 lx, in largemouth
459 bass, Howick and O’Brien 1983; lake trout, Vogel and Beauchamp 1999). At low light
460 intensities (below the critical intensity, at 1 lx), the ability of A. stuartgranti to detect more prey
461 than Tramitichromis, but at comparable distances suggests that A. stuartgranti may have superior
462 visual abilities for prey capture at these lower light intensities. This is consistent with their
463 distribution at a wider depth range than Tramitichromis and the observation of reproductive
464 behaviors in caves (Grant et al. 1987), but whether Aulonocara species possess adaptations for
465 increased sensitivity and/or visual acuity as found in known crepuscular or nocturnal teleosts
466 (reviewed in Warrant 2004; Schmitz and Wainwright 2011) requires further study.
467 The potential for binocular vision can be revealed by looking at behavioral evidence for
468 differences in the size of visual fields under different light conditions and between species.
469 While visual predators may respond differently to stimuli in different portions of their visual
470 fields (Collin 1989; McComb and Kajiura 2007; Miyazaki et al. 2011), it is detection angle that
471 is reflects the overall size of the visual field, which is defined by the size, shape, and position of
472 the eyes (Collin and Shand 2003). A. stuartgranti demonstrates a wide range of detection angles
473 at light intensities ≥ 1 lx (±90° from body axis) and an even wider range of angles in darkness (0
474 lx, ±180° from body axis). This shift is correlated with differences in behavioral parameters at 1
475 lx versus 0 lx, which are interpreted as a shift between primarily vision-mediated prey detection
476 to lateral line-mediated prey detection. Lateral line-mediated detection of prey around the body is
21
477 enabled by the more sensitive widened cranial lateral line canals that characterize Aulonocara
478 species, and by the broad distribution of canal and superficial neuromasts on the skin of the head,
479 trunk and tail, which is typical of cichlids and of most teleosts (reviewed in Webb 2014). In
480 contrast, Tramitichromis detected prey at a range of angles (±40° from body axis) that was less
481 than half of that for A. stuartgranti (±90° from body axis) at light intensities of 1-800 lx, with
482 one exception (Watson’s U2 test, P < 0.05). This suggests that Tramitichromis, but likely not A.
483 stuartgranti, uses binocular vision and depth perception to detect prey at a distance (as
484 demonstrated in other teleosts, Sivak 1978; Blanco-Vives et al. 2011; Miyazaki et al. 2011).
485 Furthermore, Tramitichromis tends to swim higher above the substrate than A. stuartgranti when
486 searching for prey in the laboratory (Schwalbe and Webb 2014). Coupled with the use of
487 binocular vision, this search strategy could explain the tendency for Tramitichromis to detect
488 benthic prey at somewhat longer distances than A. stuartgranti (Figs. 4c, d).
489 The movements of the appendages of the live prey used in this study presumably generate
490 a visual motion stimulus, and an enhanced dispersal of an odor plume (not evaluated here), in
491 addition to a hydrodynamic stimulus, which addresses the importance of multimodal integration
492 in the formulation of prey detection behavior. However, prey type predicted detection distance
493 and detection-to-strike velocity only in A. stuartgranti, which tended to strike at live prey at
494 longer detection distances and at higher velocities than for dead prey at the same light intensities
495 (12-800 lx). At 1 lx, detection distances were about one body length or less, which is within the
496 effective range of the lateral line system (Coombs 1999). Behavior is consistent with the use of
497 the lateral line system in addition to vision for detection of live prey by A. stuartgranti in full
498 light.
22
499 In contrast, in Tramitichromis, prey type did not predict any of the four behavioral
500 parameters analyzed using GLMMs, although the interaction of prey type and light intensity did
501 predict detection distance. However, the examination of data trends showed that Tramitichromis
502 demonstrates comparable values and trends for live and dead prey with reference to number of
503 prey detections, detection distance (despite the significance of its interaction with prey type) and
504 detection-to-strike velocity at light intensities of 1-800 lx. These results also substantiate results
505 of a prior laboratory study (Schwalbe and Webb 2014) that showed that Tramitichromis is a
506 visual predator, which is not dependent on the detection of hydrodynamic stimuli generated by
507 live prey. However, the lack of significance of prey type indicates that Tramitichromis does not
508 respond to a visual motion stimulus that are likely to have been generated by live (but not dead)
509 prey, which is surprising given the feeding strategies that these fish employ in nature. In the
510 relatively shallow, well-lit waters of Lake Malawi, Tramitichromis species typically capture prey
511 by plunging into the substrate, filling their mouth with sand, and sifting out prey with their gill
512 rakers (= sand sifting, Fryer 1959). The sensory basis for the plunge and sift feeding behavior
513 needs to be determined experimentally, but the results of this study suggest that it is a visual
514 stimulus and not an associated motion stimulus generated by live prey that influences where
515 Tramitichromis initiates feeding behavior in the field.
516
517 The connection between experimental light conditions and light levels in Lake Malawi
518
519 As in other lakes, the photic conditions in Lake Malawi are dynamic and many factors influence
520 the light environment, including habitat type, water depth, and proximity to the lake bottom
521 (Sabbah et al. 2011), as well as meteorological events, eutrophication, turbidity, and both diurnal
23
522 and seasonal changes in light quality and quantity. In shallow water, full spectrum light is
523 typically present and middle wavelengths transmit best, but shorter and longer wavelengths
524 attenuate rapidly (Dalton et al. 2010). Further, the irradiance spectrum differs between waters
525 overlying sandy and rocky substrates, where light transmission in water above sand is shifted to
526 longer wavelengths compared to that above rocky habitats (Sabbah et al. 2011).
527 The extent to which species of Aulonocara and Tramitichromis forage using vision at
528 different depths can be approximated by comparing behavioral data from the current study to
529 estimates of depths at which particular light intensities are predicted to occur in Lake Malawi.
530 The light extinction coefficients in Table 1 are representative of pelagic (ε = 0.10 m-1, Patterson
531 et al. 2000; ε = 0.13 m-1, Guildford et al. 2007) and nearshore (ε = 0.43 m-1, Guildford et al.
532 2007) habitats in Lake Malawi, but disparities in water clarity between these areas are likely
533 influenced by nutrient loading and sedimentation from deforestation, intense agricultural
534 practices, and erosion in nearshore areas (Bootsma and Jorgensen 2006). Estimations based on
535 low light extinction coefficients (e.g. ε = 0.10 m-1 or 0.13 m-1) suggest that Aulonocara species
536 could visually detect prey at 71 to 92 m (≥12 lx) and with some visual limitations at ~89 to 115
537 m where light levels are at ~1 lx. Some Aulonocara species are found to depths of 120 m
538 (Konings 2007), so they may be able to visually detect prey in these depths at midday when light
539 intensities are highest. Alternatively, when light extinction coefficients are used (ε = 0.43 m-1),
540 the maximum depths at which Aulonocara species could reliably detect prey are greatly reduced
541 (to 21 m and 27 m, respectively). In the lab, Tramitichromis was able detect prey at a light
542 intensity of 1 lx, which translates to depths of 89 to 115 m if the light extinction coefficient is
543 low. However, these fish are typically found in shallower waters (<15 m, Konings 1990, 2007),
544 so the ability of Tramitichromis to find prey at 1 lx is more relevant for the potential for feeding
24
545 early or late during the day. Given its dependence on vision for prey detection (Schwalbe and
546 Webb 2014), Tramitichromis species may be limited to shallow habitats so that the visual
547 detection of prey is not compromised. In contrast, Aulonocara species can feed at low light
548 intensities and in the dark, which can explain the wider range of depths at which they occur in
549 Lake Malawi. They may also be crepuscular or nocturnal in habit, which may also facilitate other
550 behaviors (e.g., social interactions) at low light intensities.
551
552 Conclusions
553
554 A. stuartgranti fed on prey at a range of ecologically relevant light intensities, including
555 darkness, and Tramitichromis was also able to feed at low light intensities, but not in darkness.
556 In A. stuartgranti, the influence of time of day on several aspects of its behavior suggests that it
557 may use circadian rhythms to regulate behavior in nature where diurnal light cues may not be
558 available (e.g. at greater depth, in caves). The integration of visual and non-visual (e.g., lateral
559 line) sensory modalities can explain the statistically non-significant trends in behavior. Similarly,
560 the dramatic change in behavior from 1 lx to 0 lx is consistent with a transition from primarily
561 vision-mediated to exclusively lateral line-mediated prey detection behavior. In contrast to A.
562 stuartgranti, Tramitichromis depends on vision-mediated prey detection (Schwalbe and Webb
563 2014); in this study its behavior was significantly affected by the interaction of light intensity
564 with time of day, but these two factors could not be teased apart. Finally, in an ecological
565 context, the tendency of Tramitichromis species to live in shallower, well-lit habitats, in contrast
566 to Aulonocara species, which live at a wide range of depths and light environments, suggests that
567 sensory capabilities may allow Aulonocara species to escape competition with Tramitichromis
25
568 species for prey resources, thus facilitating niche differentiation between these taxa. Field
569 observations in Lake Malawi are needed to test this hypothesis, which would provide an
570 important link between the morphology, feeding behavior, and ecology of cichlid fishes.
571
572 Acknowledgements
573
574 We thank Karen Carleton (University of Maryland) for her expertise in cichlid vision, Stephanie
575 Guildford (University of Minnesota Duluth) for her expertise on light levels in Lake Malawi, and
576 Graham Forrester (University of Rhode Island) for statistical expertise. We also thank Edward
577 Baker (Facilities Manager, RI NSF EPSCoR Marine Life Science Facility), Emily Becker,
578 Brandon Fuller, Christopher Holland, Callie Veelenturf, Rebecca Scott, and Benjamin Sevey for
579 assistance with fish husbandry, Christopher Kenaley (Harvard University) for doing the µCT
580 scans, and Benjamin Sevey for generating the 3D images in Fig. 1. This work was carried out
581 under an approved University of Rhode Island IACUC protocol (#AN08-11-005). This research
582 was assisted by funds provided by NSF EPSCoR Cooperative Agreement EPS-0554548 and
583 EPS-100405, the College of the Environment and Life Sciences (University of Rhode Island) and
584 was supported by NSF grant IOS 0843307 to JFW.
585
586 References
587
588 Ali MA (1959) The ocular structure, retinomotor and photobehavioral responses of juvenile
589 pacific salmon. Can J Zoology 37:965-996
26
590 Atema J (1971) Structures and functions of the sense of taste in the catfish (Ictalurus natalis).
591 Brain Behav Evol 4:273-294
592 Batty RS, Blaxter JHS, Libby DA (1986) Herring (Clupea harengus) filter-feeding in the dark.
593 Mar Biol 91:371-375
594 Bergstrom MA and Mensinger AF (2009) Interspecific resource competition between the
595 invasive round goby and three native species: logperch, slimy sculpin, and spoonhead
596 sculpin. Trans Am Fish Soc 138:1009-1017
597 Blanco-Vives B, Aliago-Guerrero M, Cañavate JP, García-Mateos G, Martín-Robles AJ,
598 Herrera-Pérez P, Muñoz-Cueto JA, Sánches-Vázquez FJ (2011) Metamorphosis induces
599 a light-dependent switch in Senegalese sole (Solea senegalensis) from diurnal to
600 nocturnal behavior. J Biol Rhythms 27:135-144
601 Bootsma H, Jorgensen SE (2005) Lake Malawi/Nyasa: Experience and lessons learned brief. pp
602 259–276. In: companion CD-ROM for: ILEC (2005). Managing lakes and their basins for
603 sustainable use: A report for lake basin managers and stakeholders. International Lake
604 Environment Committee Foundation, Kusatsu, Japan
605 Collin SP (1989) Topographic organization of the ganglion cell layer and intraocular
606 vascularization in the retinae of two reef teleosts. Vision Res 29:765-775
607 Collin SP, Shand J (2003) Retinal sampling and the visual field in fishes. In: Collin SP, Marshall
608 NJ (eds) Sensory processing in aquatic environments. Springer, New York, pp 139-169
609 Confer JL, Howick GL, Corzette MH, Kramer SL, Fitzgibbon S, Landesberg R (1978) Visual
610 predation by planktivores. Oikos 31:27-37
611 Coombs S (1999) Signal detection theory, lateral-line excitation patterns and prey capture
612 behaviour of mottled sculpin. Anim. Behav. 58: 421–430
27
613 Dalton BE, Cronin TW, Marshall NJ, Carleton KL (2010) The fish eye view: are cichlids
614 conspicuous? J Exp Biol 213:2243-2255
615 Denton EJ, Gray JAB (1988) Mechanical factors in the excitation of the lateral lines of fish. In:
616 Atema J, Fay RR, Popper AN, Tavolga WN (eds) Sensory biology of aquatic animals.
617 Springer. New York, pp 595-617
618 Denton EJ, Gray JAB (1989) Some observations on the forces acting on neuromasts in fish
619 lateral line canals. In: Coombs S, Gorner P, Münz H (eds) The mechanosensory lateral
620 ine: neurobiology and evolution. Springer, New York, pp 229-246
621 Diehl S (1988) Foraging efficiency of three freshwater fishes: effects of structural complexity
622 and light. Oikos 53:207-214
623 Dunbrack RL, Dill LM (1984) Three-dimensional prey reaction field of the juvenile coho salmon
624 (Oncorhynchus kisutch). Can J Fish Aquat Sci 41:1176-1182
625 Durham SR, Dietl GP, Visaggi CC (2012) The mismeasure of behavior: a natural history
626 revision of prey preference in the banded tulip snail. J Shellfish Res 31:101-109
627 Fryer G (1959) The trophic interrelationships and ecology of some littoral communities of Lake
628 Nyasa with especial reference to the fishes, and a discussion of the evolution of a group
629 of rock-frequenting Cichlidae. Proc Zool Soc Lond 132:153-281
630 Fryer G, Iles TD (1972) The cichlid fishes of the great lakes of Africa: their biology and
631 evolution. Oliver and Boyd, Edinburgh
632 Gardiner JM, Motta PJ (2012) Largemouth bass (Micropterus salmoides) switch feeding
633 modalities in response to sensory deprivation. Zoology 115: 78-83
634 Grant SM, Dieckhoff HW, Mayland HJ, Meyer MK (1987) Ecology of Aulonocara Regan, 1922
635 in Lake Malawi. Cour. Forsch.-Inst. Senckenberg 94:131-139
28
636 Guildford SJ, Bootsma HA, Fee EJ, Hecky RE, Patterson G (2000) Phytoplankton nutrient status
637 and mean water column light intensity in Lakes Malawi and Superior. Aquat Ecosyst
638 Health 3:35-45
639 Guildford SJ, Bootsma HA, Taylor WD, Hecky RE (2007) High variability of phytoplankton
640 photosynthesis in response to environmental forcing in oligotrophic Lake Malawi/Nyasa.
641 J Great Lakes Res 33:170-185
642 Harden Jones, FR (1956) The behaviour of minnows in relation to light intensity. J Exp Biol
643 33:271-281
644 Henderson MA, Northcote TG (1985) Visual prey detection and foraging in sympatric cutthroat
645 trout (Salmo clarki clarki) and dolly varden (Salvelinus malma). Can J Fish Aquat Sci
646 42:785-790
647 Hofmann CM, O’Quin KE, Marhsall NJ, Cronin TW, Seehausen O, Carleton KL (2009) The
648 eyes have it: regulatory and structural changes both underlie cichlid visual pigment
649 diversity. PLOS Biol 7(12):e1000266
650 Holbrook SJ, Schmitt RJ (1984) Experimental analyses of patch selection by foraging black
651 surfperch (Embiotoca jacksoni Agazzi). J Exp Mar Biol Ecol 79:39-64
652 Howick GL, O’Brien WJ (1983) Piscivorous feeding behavior of largemouth bass: experimental
653 analysis. Can J Fish Aquat Sci 42:785-790
654 Janssen J (1997) Comparison of response distance to prey via the lateral line in the ruffe and
655 yellow perch. J Fish Biol 51:921-930
656 Jordan R, Kellogg K, Howe D, Juanes F, Stauffer J, Loew E (2006) Photopigment spectral
657 absorbance of Lake Malawi cichlids. J Fish Biol 68:1291-1299
29
658 Kirk JTO (2011) Light and photosynthesis in aquatic ecosystems, 3rd edn. Cambridge University
659 Press, New York
660 Konings A (1990) Koningʼs book of cichlids and other fishes of Lake Malawi. TFH Publications
661 Inc., Neptune City, New Jersey
662 Konings A (2007) Malawi cichlids in their natural habitat, 4th edn. Cichlid Press, El Paso, Texas
663 Laur DR, Ebeling AW (1983) Predator-prey relationships in surfperches. Env Biol Fish 8:217-
664 229
665 Lythgoe JN (1979) The ecology of vision. Clarendon Press, Oxford
666 Mazur MM, Beauchamp DA (2003) A comparison of visual prey detection among species of
667 piscivorous salmonids: effects of light and low turbidities. Environ Biol Fish 67:397-405
668 McComb DM, Kajiura SM (2008) Visual fields of four batoid fishes: a comparative study. J Exp
669 Biol 211:482-490
670 McWilliam RA, Minchinton TE, Ayre DJ (2013) Despite prolonged association in closed
671 populations, an intertidal predator does not prefer abundant local prey to novel prey. Biol
672 J Linn Soc 108:812-820
673 Miyazaki T, Iwami T, Meyer-Rochow VB (2011) The position of the retinal area centralis
674 changes with age in Champsocephalus gunnari (Channichthyidae), a predatory fish from
675 coastal Antarctic waters. Polar Biol 34:1117-1123
676 Moller P (2002) Multimodal sensory integration in weakly electric fish: a behavioral account. J
677 Physiology-Paris 96: 547-556
678 Montgomery JC, Bleckmann H, Coombs S (2014) Sensory ecology and neuroethology of the
679 lateral line. In: Coombs S, Bleckmann H, Fay RR, Popper AN (eds) The lateral line
680 system. Springer, New York, pp 121-150
30
681 Montgomery JC, Coombs S (1992) Physiological characterization of lateral line function in the
682 Antarctic fish Trematomus bernacchii. Brain Behav Evol 40:209-216
683 Montgomery JC, Diebel C, Halstead MBD, Downer J (1999) Olfactory search tracks in the
684 Antarctic fish Trematomus bernacchii. Polar Bio 21:151-154
685 Montgomery JC, McDonald F, Baker CF, Carton AG, Ling N (2003) Sensory integration in the
686 hydrodynamic world of rainbow trout. Roy Soc Lond B Bio 270(2): S195-S197
687 Montgomery JC, Windsor S, Basset D (2009) Behavior and physiology of mechanoreception:
688 separating signal and noise. Integr Zool 4:3-12
689 Parry JWL, Carleton KL, Spady T, Carboo A, Hunt DM, Bowmaker JK (2005) Mix and match
690 color vision: tuning spectral sensitivity by differential opsin gene expression in Lake
691 Malawi cichlids. Curr Biol 15:1734-1739
692 Partridge BL, Pitcher TJ (1980) The sensory basis of fish schools: relative roles of lateral line
693 and vision. J Comp Physiol 135: 315-325
694 Parzefall J (1993) Behavioural ecology of cave-dwelling fishes. In: Pitcher T (ed) Behaviour of
695 teleost fishes, 2nd edn. Chapman & Hall, London, pp 573-608
696 Patterson G, Hecky RE, Fee EJ (2000) Effect of hydrological cycles on planktonic primary
697 productivity in Lake Malawi/Niassa. Adv Ecol Res 31:421-430
698 Richmond HE, Hrabik TR, Mensinger AF (2004) Light intensity, prey detection and foraging
699 mechanisms of age 0 year yellow perch. J Fish Biol 65:195-205
700 Rickel A, Genin A (2005) Twilight transitions in coral reef fish: the input of light-induced
701 changes in foraging behaviour. Anim Behav 70:133-144
702 Ryer CH, Olla BL (1999) Light-induced changes in the prey consumption and behavior of two
703 juvenile planktivorous fish. Mar Ecol Prog Ser 181:41-51
31
704 Sabbah S, Gray SM, Boss ES, Fraser JM, Zatha R, Hawryshyn CW (2011) The underwater
705 photic environment of Cape Maclear, Lake Malawi: comparison between rock- and sand-
706 bottom habitats and implications for cichlid fish vision. J Exp Biol 214:487-500
707 Schmidt D, O’Brien WJ (1982) Planktivorous feeding ecology of arctic grayling (Thymallus
708 arcticus). Can J Fish Aquat Sci 39:475-482
709 Schmitt R, Holbrook SJ (1984) Ontogeny of prey selection by black surfperch Embiotoca
710 jacksoni (Pisces: Embiotocidae): the roles of fish morphology, foraging behavior, and
711 patch selection. Mar Ecol Prog Ser 63:6-12
712 Schmitz L, Wainwright PC (2011) Nocturnality constrains morphological and functional
713 diversity in the eyes of reef fishes. BMC Evol Biol 11:338
714 Schwalbe MAB, Bassett DK, Webb JF (2012) Feeding in the dark: lateral-line-mediated prey
715 detection in the peacock cichlid Aulonocara stuartgranti. J Exp Biol. 215:2060-2071
716 Schwalbe MAB, Webb, JF (2014) Sensory basis for detection of benthic prey in two Lake
717 Malawi cichlids. Zoology 117:112-121
718 Sivak JG (1978) The functional significance of the aphakic space of the fish eye. Can J Zool
719 56:513-516
720 Taplin RH (2007) Experimental design and analysis to investigate predator preferences for prey.
721 J Exper Mar Biol Ecol 344:116-122
722 Townsend CR, Risebrow AJ (1982) The influence of light level on the functional response of a
723 zooplanktonivorous fish. Oecologia 53:293-295
724 Vinyard GL, O’Brien WJ (1976) Effects of light and turbidity on reaction distance of bluegill
725 (Lepomis macrochirus). J Fish Res Board Can 33:2845-2849
32
726 Vogel JL, Beauchamp DA (1999) Effects of light, prey size, and turbidity on reaction distances
727 of lake trout (Salvelinus namaycush) to salmonid prey. Can J Fish Aquat Sci 56:1293-
728 1297
729 Warrant EJ (2004) Vision in the dimmest habitats on Earth. J Comp Physiol A 190:765-789
730 Webb JF (2014) Morphological diversity, development, and evolution of the mechanosensory
731 lateral line system. In: Coombs S, Bleckmann H, Fay RR, Popper AN (eds) The lateral
732 line system. Springer, New York, pp 17-72
733 Webb JF, Montgomery JC, Mogdans J (2008) Bioacoustics and the lateral line system of fishes.
734 In: Webb JF, Fay RR, Popper AN (eds) Fish Bioacoustics. Springer, New York, pp 145–
735 182
736
33
737 Figure Legends
738
739 Fig. 1 Lateral line canals and canal neuromasts in Aulonocara stuartgranti (widened canals) and
740 Tramitichromis (narrow canals) visualized using fluorescent vital staining (4-di-2-ASP, 63 µM, 5
741 min; a, c), µCT imaging (reconstructed from 14 µm slices; b, d, e, g, h), and scanning electron
742 microscopy (f). a Lateral view of A. stuartgranti revealing series of larger infraorbital (IO),
743 mandibular (MD), and preopercular (PO) canal neuromasts and very small superficial
744 neuromasts on the skin (juvenile, 25 mm standard length [SL]). Neuromast number and
745 distribution is the same in Tramitichromis b µCT reconstruction of A. stuartgranti (adult, 78 mm
746 SL) indicating the location of the supraorbital (SO), IO, MD, and PO canals in dermatocranial
747 bones. c Ventral view of the head of A. stuartgranti (juvenile, 28 mm SL), revealing canal
748 neuromasts in the MD and PO canals. d A. stuartgranti (adult, 78 mm SL) and e Tramitichromis
749 (adult, 79 mm SL) in ventral view. Asterisks (*) denote the locations of the MD and PO canal
750 neuromasts, as visualized in c; canal neuromasts are found in floor of the canal, between canal
751 pore positions in the canal roof. Note the much larger pores in A. stuartgranti (d) than in
752 Tramitichromis (e). f MD canal neuromast in a juvenile A. stuartgranti. Ciliary bundles of the
753 sensory hair cells are evident in an elongate sensory strip in the middle of the diamond-shaped
754 neuromast. Double-headed arrow below the sensory strip indicates the axis of physiological
755 sensitivity of the hair cells, as well as the long axis of the canal in which the neuromast is found.
756 Scale bar = 10 µm. g A. stuartgranti and h Tramitichromis in frontal-ventral view with the pores
757 of the SO, IO, MD, and PO canals that are directed ventrally, toward the source of stimuli
758 generated by benthic prey. The pores on the right side of each fish in g and h have been
759 enhanced to increase their visibility.
760
34
761 Fig. 2 Experimental setup and light conditions used to record feeding behavior of A. stuartgranti
762 and Tramitichromis. a Diagram of experimental tank with front panel of light curtain removed.
763 Light sources originate from two fluorescent light fixtures (full spectrum light) and two infrared
764 (IR) lights. b Sample light spectra generated by fluorescent lights in behavioral trials. The y-axis
765 was compressed to illustrate peaks in the visual spectrum (400-700 nm) that occurred in 1-800 lx
766 trials. These peaks were consistent when light intensity was decrease with the addition of neutral
767 density filters among trials (see Table 1). The peak at 840 nm is from two IR lights in 0 and 1 lx
768 trials only. c Mean (± s.e.m.) light intensities measured before and after trials indicating that light
769 intensity did not differ before and after trials at any of the light intensities used (students t-test, P
770 > 0.05)
771
772 Fig. 3 Total number of prey detections by prey type (white bars = live tethered brine shrimp,
773 grey bars = dead tethered brine shrimp) for a A. stuartgranti (n = 6 fish) and b Tramitichromis (n
774 = 6 fish) at five different light intensities. Maximum number of possible prey detections = 72 for
775 each light intensity 12 776 12
777 Fig. 410 Three behavioral parameters defining prey detection in A. stuartgranti (n = 6 fish) and 10 778 Tramitichromis (n = 6 fish) at five different light intensities. (a, c, e) Mean (± s.e.m.) number of 8 8 779 prey detections (maximum 6 live, 6 dead tethered brine shrimp), Aulonocaradetection distance, Live and Aulonocara Live 6 Aulonocara Dead 780 detection-to-strike velocity, for6 A. stuartgranti feeding on live ( ) and dead ( )Aulonocara prey, and Dead Tramitichromis Live Tramitichromis Live 781 (b, d, f4) Tramitichromis feeding on live ( ) and dead ( Tramitichromis) prey Dead 4 Tramitichromis Dead Detection Distance[cm] Detection 782 2 Distance[cm] Detection 2
0 0 0.001 0.1 10 1000 0.001 0.1 10 1000 Light Intensity [lx] Light35 Intensity [lx]
783 Fig. 5 Frequency of prey strikes (live and dead prey combined, 12 total prey/trial) during glide or
784 pause phases of swimming at five different light intensities in a A. stuartgranti (n = 6 fish) and b
785 Tramitichromis (n = 6 fish)
786
787 Fig. 6 Detection angle for live and dead prey combined (=12 total prey/trial) at five different
788 light intensities for a A. stuartgranti (n = 6 fish) and b Tramitichromis (n = 6 fish). Black lines
789 represent the proportion of prey detections grouped into 20° intervals. Fish snout is at the center
790 and fish is facing 0° (indicated by the grey arrow in the top plot in a). The thin line represents the
791 mean angle for all trials. Results for Tramitichromis at 0 lx were not included here due to the
792 small number of strikes (n = 3 strikes)
793
794 Fig. 7 Mean (± s.e.m.) prey preference scores (following Taplin 2007) for a A. stuartgranti (n =
795 6 fish) and b Tramitichromis (n = 6 fish) feeding on six live (white bars) and six dead (gray bars)
796 tethered adult brine shrimp in trials at five different light intensities. Preferences scores were
797 calculated by taking the mean of the rank order in which prey were captured. The dotted line (=
798 6.5) indicates the mean preference score with no preference for either prey type. Scores <6.5
799 (below dotted line) indicate a preference. Significantly different preference scores between live
800 and dead prey indicated by an asterisk (*, paired t-test, P < 0.05, Table 5)
36
Table 1 The relationship of measured light intensity (mean lux and PAR, ± s.e.m. measured immediately after behavioral trials) and predicted depths at which these intensities occur in Lake Malawi. Calculations were based on midday sunlight levels, three light extinction coefficients ( 0.10 m-1, Patterson et al. 2000; 0.13 m-1, 0.43 m-1, Guildford et al. 2007), and light intensities under natural conditions (Harden Jones 1955, Ali 1959). Light intensities were achieved by varying the height of two fluorescent fixtures (ballasts) and/or covering these fixtures with several neutral density filters
Light Intensity Light Extinction Coefficient
Lux PAR ε = 0.10 m-1 ε = 0.13 m-1 ε = 0.43 m-1
Lumen/m2 µmol photons/m2/s Depth (m) Depth (m) Depth (m) Light intensities under natural conditions
800 (800.8 ± 5.4) 11.0 (11.0 ± 0.10) 52 40 12 Very cloudy day
112 (112.4 ± 1.9) 1.5 (1.51 ± 0.03) 72 55 17 Twilight
12 (12.0 ± 0.3) 0.2 (0.16 ± 0.01) 92 71 21 Twilight
1 (1.4 ± 0.1) 0.03 (0.03 ± 0.01) 115 89 27 Full moon/deep twilight
0 (0 ± 0.1) 0 (0.000 ± 0.003) NA NA NA New moon Table 2 Determination of GLMM types used to analyze four parameters of feeding behavior at five different light intensities (0-800 lx) in interspecific and intraspecific comparisons
Source Distribution Link Covariance AICC
Structure
4-way GLMMs
Number of prey strikes Multinomial Negative log-log AR(1) 1,713.0
Detection distance Normal* Identity AR(1) 177.9
Detection-to-strike velocity Normal* Identity AR(1) -32.4
Swimming phase at prey detection Binomial Probit AR(1) 1932.8
3-way GLMMs
Aulonocara stuartgranti
Number of prey strikes Multinomial Probit AR(1) 698.1
Detection distance Normal* Identity AR(1) 44.9
Detection-to-strike velocity Normal* Identity AR(1) -24.6
Swimming phase at prey detection Binomial Probit AR(1) 1065.9
Tramitichromis
Number of prey strikes Multinomial Negative log-log AR(1) 955.4
Detection distance Normal* Identity AR(1) 130.6
Detection-to-strike velocity Normal* Identity AR(1) -7.2
Swimming phase at prey detection Binomial Probit AR(1) 869.1
Note: the table includes information on the error distribution and link function. The first-order auto-regressive process [AR(1)] was used for the covariance structure in all models. The most parsimonious model was selected based on the corrected Akaike information criterion (AICC).
*Data was log10 transformed to achieve normality (normality assessed with Kolmogorov-Smirnov test). Table 3 Summary of 4-way GLMM statistics for prey detection behavior for two species (A. stuartgranti, n = 6 fish; Tramitichromis, n = 6 fish) feeding on two prey types (live, dead) at five light intensities (0-800 lx). Only those factors that are significant for at least one behavioral parameter are listed. See Table 2 for details of GLMMs used
Number of prey Detection-to-strike Swimming phase at prey Detection distance detections velocity detection
Source F (df) P value F (df) P value F (df) P value F (df) P value
Species (S) n.s. n.s. n.s. n.s.
Light intensity (L) 14.390 (1,99) <0.001 9.480 (1,481) 0.002 8.919 (1,511) 0.003 8.276 (1,18) 0.010
Time of day (T) 22.203 (1,99) <0.001 17.342 (1,512) <0.001 16.838 (1,513) <0.001 0.038 (1,129) 0.847
Prey type (P) 4.549 (1,99) 0.035 1.145 (1,503) 0.285 3.876 (1,504) 0.050 3.037 (1,514) 0.082
S × L 8.950 (1,99) 0.004 n.s. n.s. n.s.
S × P n.s. n.s. n.s. 4.288 (1,514) 0.039
L × T 14.101 (1,99) <0.001 10.479 (1,482) 0.001 9.957 (1,512) 0.002 6.044 (1,62) 0.017
S × L × T 10.464 (1,99) 0.002 n.s. n.s. n.s.
n.s. = not significant (P > 0.05)
P values < 0.05 are shown in bold
Table 4 Summary of 3-way GLMM statistics for prey detection behavior for A. stuartgranti (n =
6 fish) and Tramitichromis (n = 6 fish) feeding on two prey types (live, dead) at five different light intensities (0-800 lx). See Table 2 for details of GLMMs used
Aulonocara stuartgranti Tramitichromis
Model term F (df) P value F (df) P value
Number of prey detections
Light intensity (L) 0.282 (1,47) 0.598 11.867 (1,47) 0.001
Time of day (T) 11.212 (1,47) 0.002 23.887 (1,47) <0.001
Prey type (P) 1.649 (1,47) 0.205 0.109 (1,47) 0.743
L × T 0.293 (1,47) 0.591 12.162 (1,47) 0.001
L × P 0.845 (1,47) 0.363 0.078 (1,47) 0.780
T × P 0.968 (1,47) 0.330 0.003 (1,47) 0.956
L × T × P 0.675 (1,47) 0.416 0.037 (1,47) 0.847
Detection distance
Light intensity (L) 2.772 (1,242) 0.097 6.185 (1,217) 0.014
Time of day (T) 26.812 (1,291) <0.001 5.655 (1,223) 0.018
Prey type (P) 8.220 (1,286) 0.004 0.000 (1,217) 0.986
L × T 3.408 (1,240) 0.066 6.677 (1,217) 0.010
L × P 0.002 (1,286) 0.965 4.019 (1,220) 0.046
T × P 4.604 (1,286) 0.033 0.026 (1,217) 0.872
L × T × P 0.004 (1,286) 0.949 4.211 (1,220) 0.041
Detection-to-strike velocity
Light intensity (L) 3.158 (1,200) 0.077 4.695 (1,223) 0.031
Time of day (T) 16.895 (1,290) <0.001 8.058 (1,221) 0.005
Prey type (P) 20.107 (1,286) <0.001 0.012 (1,218) 0.912
L × T 3.465 (1,198) 0.064 5.495 (1,223) 0.020 L × P 0.385 (1,286) 0.535 2.220 (1,219) 0.138
T × P 14.330 (1,286) <0.007 0.014 (1,218) 0.907
L × T × P 0.393 (1,286) 0.531 2.300 (1,219) 0.131
Swimming phase at prey detection
Light intensity (L) 0.289 (1,291) 0.592 3.208 (1,223) 0.075
Time of day (T) 1.593 (1,291) 0.208 0.068 (1,223) 0.794
Prey type (P) 0.147 (1,291) 0.701 3.794 (1,223) 0.053
L × T 0.128 (1,291) 0.720 2.873 (1,223) 0.091
L × P 0.220 (1,291) 0.639 0.433 (1,223) 0.511
T × P 0.202 (1,291) 0.653 2.588 (1,223) 0.109
L × T × P 0.211 (1,291) 0.646 0.438 (1,223) 0.509
P values < 0.05 are shown in bold
Table 5 Summary of the paired t-tests comparing prey preference scores of live versus dead prey of A. stuartgranti and Tramitichromis (following Taplin 2007) by light intensity
Light intensity (lx) T (df) P value
Aulonocara stuartgranti
0 -2.853 (5) 0.036
1 -0.618 (5) 0.564
12 -1.395 (5) 0.222
112 -6.102 (5) 0.002
800 -1.892 (5) 0.117
Tramitichromis
0 -0.797 (5) 0.461
1 -2.396 (5) 0.062
12 -1.379 (5) 0.226
112 -2.441 (5) 0.059
800 -13.647 (5) <0.001
P values < 0.05 are shown in bold